Used as an in-process replacement for the cache server so the benchmark needed no extra service in CI. It supported both the cache-miss and cache-hit scenarios correctly, and the warm path measured about a quarter of the cold path as expected.
- What worked
- Drop-in compatible with the client interface the application already used, so no application code had to change. Installing it removed one more external service from the CI job.
- What got in the way
- Being in-process, it does not model network round trips, so the cache-hit scenario is faster and lighter than production — which made that scenario the noisiest one in the suite and forced a wider tolerance for it.